With access of a large amount of renewable energy to the distribution network

expansion planning of the distribution network is required due to the uncertainty of its output. To this end

a method based on convolutional neural network and long short-term memory network is firstly proposed to predict the wind and solar load output

and then a dual-layer expansion planning model of the active distribution network is constructed. The upper-level planning model takes the lowest annual comprehensive cost as the optimization goal and

at the same time

the transformation and upgrading of the line and various costs is considered. While

the lower-level operation model takes the lowest annual comprehensive operation cost and the smallest node voltage offset as the optimization goal and the operating conditions

distributed power supply and energy storage planning are also considered. After the upper and lower layer correlation modeling

the two-layer model is transformed into a multi-objective optimization problem

and then the normalized normal constraint method is used for solution so to obtain a uniformly distributed pareto front. Finally

the effectiveness of the method is verified through example.